A Robotic Bone Drilling Methodology Based on Position ...vigir.missouri.edu/~gdesouza/Research/Conference_CDs/BioRob_20… · A Robotic Bone Drilling Methodology Based on Position
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
A Robotic Bone Drilling Methodology Based on Position Measurements
Marcos Louredo, Inaki Dıaz and Jorge Juan Gil
Abstract— Over the past decades many studies have dealtwith the development of robotic tools to improve the processof bone drilling. The main difficulty of the operation resides inthe ability to detect bone layer transitions and/or protrusionsduring the procedure so that damage to surrounding tissue isminimized. The present paper set up a test bench in order tostudy some of the most relevant drilling methodologies in theliterature. The study illustrates some of the drawbacks, and itproposes a new drilling methodology that provides improvedresults.
Index Terms— Assisted Surgery, Bone Drilling
I. INTRODUCTION
The present work focuses on bone-machining surgical
procedures such as drilling, reaming and sawing, all very
common in different surgical interventions. Specifically, our
work focuses on bone-drilling procedures.
In any hospital around the world, bone drilling procedures
are performed many times per day, and in very different
surgical specialties such as orthopedic surgery, ear surgery,
maxillofacial surgery, neurosurgery, among many others.
During the drilling procedure, the surgeon uses a drill or
a similar device to make a hole through the bone. In many
cases this operation is extremely delicate and requires great
precision and accuracy in order to drill only the desired
depth. Even the slightest deviation in the drilling path can
damage the tissue surrounding the bone (veins, arteries, brain
tissue, spinal cord, etc.) causing irreversible damage to the
patient.
Currently, drilling processes are usually carried out by em-
ploying manual electric or pneumatic drilling tools (Fig. 1).
The operation mode of these tools is very simple and similar
to the drilling tools used at home to hang a picture. The
surgeon can control the rotation speed of the drill bit (with a
pedal, button, etc.) while he/she exerts certain force against
the bone to make the hole.
The main disadvantage of such drilling tools is that there
is no way to guess when the hole is done or the desired
depth is reached. Moreover, a breakthrough can push the
drill bit further along the drilling axis due to the inertia
of the drilling force. While this undesired effect may not
be very important when drilling a wall at home, it can be
of critical relevance when drilling a bone since surrounding
tissue can be seriously damaged. Currently, the only way of
efficiently stopping the drilling procedure at the desired depth
is the experience and intuition of the surgeon. Therefore,
The authors are with the Applied Mechanics Department, CEIT, PaseoManuel Lardizabal 15, E-20018 San Sebastian (Guipuzcoa), Spain (e-mails: {mlouredo,idiaz,jjgil}@ceit.es) and with TECNUN,University of Navarra, Paseo Manuel Lardizabal 13, E-20018 San Sebastian(Guipuzcoa), Spain.
Fig. 1. Bone drilling tool.
any means of assisting the surgeon during the operation can
decrease the potential for error or mishap.
Over the last decade many solutions have been proposed
to improve the current art of surgical drilling. A number
of solutions rely on image-based trajectory control, which
involves the surgeon using X-ray images from time to time
to “see” the penetration depth. This is the current approach
taken by surgeons when depth control is critical.
Developments in the drilling tool have included mecha-
tronic systems that control either the linear or the rotational
movements of the drill bit, or both movements simultane-
ously, such that the surgeon only has to place the system
at the correct position and orientation. These semi-automatic
and automatic solutions vary depending on the methodology
they follow to control the penetration depth into the bone:
(i) by using predefined penetration depth values [1], or (ii)
by using control algorithms that analyze the measurements
of different sensors coupled to the drill bit [2], [3], [4].
The present work focuses on the automatic drilling tools
group. Both the rotation and the linear movement of the
drill bit are automatic; that is, the surgeon places the drill
bit at the desired position and orientation and pushes the
start button. Afterward, the system carries out the drilling
procedure automatically and stops the drill bit either at the
layer transition or at bone breakthrough, according to the
surgeon’s requirements. An optimal control methodology
should be able to move the drill bit along its trajectory in
order to achieve a minimum level of protrusion of the drill
bit beyond the desired point.
This paper provides a review of mechatronic drilling
methodologies found in the literature. Moreover, it analyzes
some of the most relevant algorithms in a test bench specif-
The Fourth IEEE RAS/EMBS International Conferenceon Biomedical Robotics and BiomechatronicsRoma, Italy. June 24-27, 2012
The control unit that drives the test bench consists of a
control box and a display. All the electronics are placed
in this module. It receives sensor measurements from the
drilling guide, analyzes them and controls the drilling pro-
cess. Specifically, for this prototype, a dSpace DS1104
control board is used, and control algorithms run at a fixed
sampling rate of 1 kHz.
B. Experiments
Drilling experiments were performed on cortical bones
along the middle section of cooked bovine femoral shafts.
The whole femur, which was stripped of all soft tissue, was
clamped rigidly onto a bone holder in order to ensure that
there was no system compliance. The input feed rate for the
axial movement was set to 12 mm/min, and the rotational
speed of the drill bit was 7500 rev/min. A 6 mm diameter
surgical drill bit was used for the experiments.
Notice that some limitations, which do not come into play
in real drilling processes, are taken into consideration: (i) the
bone is rigidly attached to a structure (rigid bone holder), (ii)
no refrigeration is applied, and (iii) cooked animal bones
are used. Although these considerations limit application
to a real environment, they still allow the basic working
principle of the detection algorithm to be validated and allow
a comparison with prior state of the art that also considered
similar testing conditions [8], [9], [12].
Fig. 5 shows a picture of the final test bench set-up. Three
representative drilling methodologies of the related work
were implemented on it, and results are described next. We
refer the reader to the references cited for each methodology
for further details about the algorithms applied.
Fig. 6 shows the data obtained when drilling with the
detection methodology presented by Ong and Bouazza-
Marouf [9]. The authors used the Force Difference between
1157
Fig. 5. Test bench set-up and bone holder.
Successive Samples (FDSS), filtered by a modified Kalman
estimator (K-FDSS), to detect bone layer transitions. This
detection occurs when the K-FDSS signal passes through
a certain threshold value (0.001 N in the experiment). The
Force signal plotted in the figure is the penetrating force
measured by the sensor in the axial direction.
Fig. 6. Drilling procedure on two cortical bone layers using the methodby Ong and Bouazza-Marouf [9]. Dots 1 and 3 represent the imminent drillbit breakthrough, while squares 2 and 4 show the detection points providedby the method. Triangles show false detection points.
It can be seen from the figure that the implementation of
this methodology for detecting drill bit breakthrough results
in late detection. Moreover, a sensitive threshold value (such
as the one applied in the experiment) generates many false
detection points before the drilling begins (t = 0 − 14 s),
while less sensitive values imply higher delays in the detec-
tion point.
Fig. 7 shows the data obtained when drilling with the
wavelet-based detection methodology presented by Colla and
Allota [16]. Using a tree-structured wavelet decomposition
stage, the authors obtained three useful output sequences: the
approximation coefficients at the first decomposition level
and two detail coefficients (d1 and d2). These latter two
sequences present a peak corresponding to each breakdown
in the original signal. These peaks can be easily detected by
a simple threshold comparator (γ2 and γ13).
By using a threshold value of γ13, protrusion at the first
Fig. 7. Drilling procedure of two cortical layers using the method by Collaand Allota [16]. Dots represent the imminent drill bit breakthrough, whilesquares show the detection points provided by the method.
cortical wall is detected on time, however, the detection for
the second cortical wall occurs when the breakthrough is
almost done. Again, tuning the threshold value improves
some precision, but it leads to other undesired effects such
as false detection points.
The last method analyzed at the test bench is the one
proposed by Kaburlasos et al. [18] (Fig. 8). The authors
used neural network theory and fuzzy logic to identify four
drilling stages (Sa, Sb, Sc, Sd). Transition from one stage to
another is held at a representative drilling point that allows
the procedure to be controlled: 1) bone drilling starts, 2)
drilling is stabilized (penetration is greater than half the
radius of the drill bit), 3) protrusion starts, and 4) complete
breakthrough occurs.
Fig. 8. Drilling procedure with one cortical layer using the method byKaburlasos et al. [18]. Red squares represent the transition points fromstage to stage, and the yellow dot shows the beginning of the protrusion.
The method performs well if the drilled bone has similar
thickness and homogeneity to the bones used to implement
the neural network. But again, the method lacks precision
when bone type changes or the threshold value is modified.
To sum up, the methods analyzed have some elements
in common: (i) the use of a velocity control approach for
the axial movement of the drill bit, (ii) the use of a force
sensor to measure penetration force and to feed the detection
1158
control algorithm, and (iii) the use of threshold values to
determine layer transitions. As a result, these methods have
similar limitations in detecting layer transitions just before
protrusion: (i) the signal of the force sensor is very noisy
for control purposes, and certain delay is introduced after
filtering this signal, (ii) the use of threshold values decreases
the effectiveness of the algorithms among different bone
types; moreover, depending on the sensitivity of the threshold
value it results in false detections with conservative values
and a lack of detection with wider margins.
IV. PROPOSED BONE DRILLING METHODOLOGY
The present work proposes a new control methodology for
detecting layer transitions to improve the point of detection
[21]. Our focus is on finding a method that allows fast
detection (just before protrusion), and whose effectiveness is
not conditioned by the sensitivity of using threshold values
(low false detection rate).
The proposed methodology uses a different approach to
control the linear movement along the penetration axis com-
pared to traditional velocity control. The diagram in Fig. 9
shows the axial movement control law applied to the linear
guide through the DC motor and the cable transmission. The
input to the system is a ramp-style position signal, with the
slope being equal to the desired translational drilling speed.
Then, a proportional controller is applied to the difference
between the reference position and the X position measured
by the optical encoder attached to the motor.
Fig. 9. Control diagram for the axial displacement along the guide.
Before reaching the bone, the system moves smoothly
following the speed imposed by the slope of the ramp. Once
drilling begins, the position error greatly increases due to the
resistance encountered when the drilling movement comes up
against the bone’s stiffness, and also because the controller
saturates the actuation to a proper motor torque level. Note
that once the motor is saturated, the control scheme works
as a constant force input to the system.
Simultaneously, a detection algorithm is running at the
control unit. Its duty is to predict the exact moment just
prior to the protrusion in order to stop the axial movement
of the drill. To that end, the detection algorithm monitors the
same position error signal fed to the controller in Fig. 9.
Fig. 10 shows the position reference signal, the X position
measured by the encoder, and the error signal between both
of them when drilling a bone with this control scheme. At
time t = 0 s, the drill comes into contact with the bone
and starts the process. Due to the stiffness of the bone and
the saturation of the motor, the X position lags behind the
position reference, and the error signal increases as drilling
time goes on. This situation remains unaltered until the bone
protrusion is about to happen (t ≈ 125 s).
Fig. 10. Position and error signals measured during the drilling process(at t = 0 s bone drilling starts, at t ≈ 125 s bone protrusion starts).
Just before protrusion the remaining width of the bone
layer becomes very tiny and bone stiffness is considerably
reduced. At that moment, and according to the implemented
control movement law, the system will respond by acceler-
ating while trying to follow the reference position, thereby
minimizing the position error. The implemented detection
method targets this sudden acceleration to discriminate a
bone layer transition.
In terms of the error signal, this variation is seen as a
change in the sign of its slope, from positive to negative.
This condition is implemented in the detection algorithm,
imposing the detention of the axial movement of the drill as
a response to a sign change in the slope of the error signal.
A very important feature of the proposed algorithm is that
at detection time a very thin bone layer still remains, which
the surgeon can easily break manually. This condition allows
safer drilling in critical places since the system stops the drill
just before any surrounding tissue that is different to bone is
reached. In contrast, the prior state of the art, mostly detect
layer transitions once bone protrusion occurs.
The detection algorithm successfully detects the bone
protrusion in time by employing only the measurement of
the optical encoder that is attached to the feeding motor.
This measurement, unlike a force sensor or an accelerom-
eter, has the advantage of having very low noise. Another
condition for the successful implementation of the algorithm
is the cable transmission used to feed the carriage. Its
high reversibility allows the control scheme in Fig. 9 to
be implemented, and it also allows the system to accelerate
before protrusion (Fig. 10).
20 consecutive drilling experiments were performed using
the proposed method at the test bench, and a comparison with
a detection methodology based on the force measure was also
carried out. All of them performed very similar results. The
proposed new method detected protrusion on cortical walls
earlier than force-based methods. Fig. 11 shows data from
one of the experiments (the system was commanded to detect
the protrusion but to go on drilling in order to get the data).
1159
Time (s)
X(m
m)
err
X(m
m)
X(m
m)
ref
Forc
e(N
)
Fig. 11. Drilling experiment using our proposed method. Yellow dotsrepresent the detection points found by force-signal-based methodologies,and red dots the detection points obtained by the proposed methodology.
Fig. 12 shows the holes made with the new detection
methodology (left) and with the force signal-based algo-
rithms (right) when commanding the system to stop at bone
protrusion. It can be seen how the new method is able to
stop the procedure just before the breakthrough whilst other
methods are not so accurate.
Fig. 12. Holes made with force-signal-based (left) and with the proposedmethodology (right).
V. CONCLUSIONS
This work reviewed previous related work on mechatronic
bone drilling systems and methodologies. Moreover, a test
bench was used to show some common drawbacks present
in such methods. Afterward, we described and tested a new
bone drilling methodology that improves the previous art.
The proposed drilling method avoids the use of force
sensors or accelerometers, and the detection algorithm relies
uniquely on the measure of the linear movement of the drill
bit. By using the position signal of the drill bit, better results
can be obtained than with previous systems. The position
signal is less affected by electromagnetic noise, and the
control methodology used provides earlier detection of the
layer transitions and breakthroughs.Future work will focus on an approach close to the
real conditions found in an operating theatre, taking into
account the limitations considered in this paper, and on the
development of a hand-held mechatronic drilling tool that
can take advantage of the drilling methodology proposed.
REFERENCES
[1] W. Anderson, “Depth controllable and measurable medical driverdevices and methods of use,” Patent, 2009, US 2009/0326537 A1.
[2] B. Allotta, “Surgical drill with bit penetration control and breakthroughdetection,” Patent, 2000, US 6033409.
[3] Y.-L. Hsu, S.-T. Lee, C.-F. Wang, J.-W. Chen, H.-W. Lin, and T.-C. Huang, “Automatic bone drilling apparatus for surgery operation,”Patent, 2002, US 6336931.
[4] A. A. Carl, J. Adams, K. C. Craig, D. C. Lavery, G. Fischer, S. R.Anthony, J. L. Hurst, A. K. Modi, and D. Digiulio, “Methods andsystems for controlling the operation of a tool,” Patent, 2005, US2005/0116673 A1.
[5] P. N. Brett, D. A. Baker, L. Reyes, and J. Blanshard, “An automatictechnique for micro-drilling a stapedotomy in the flexible stapesfootplate,” Proc Inst Mech Eng H, vol. 209, no. 4, pp. 255–262, 1995.
[6] D. Baker, P. Brett, M. Griffiths, and L. Reyes, “A mechatronic drillingtool for ear surgery: A case study of some design characteristics,”Mechatronics, vol. 6, no. 4, pp. 461–477, 1996.
[7] ——, “Surgical requirements for the stapedotomy tool: data and safetyconsiderations,” in 18th Annual Int. Conf. of the IEEE Engineering inMedicine and Biology Society, vol. 1, 1996, pp. 214–215.
[8] B. Allotta, F. Belmonte, L. Bosio, and P. Dario, “Study on a mecha-tronic tool for drilling in the osteosynthesis of long bones: Tool/boneinteraction, modeling and experiments,” Mechatronics, vol. 6, no. 4,pp. 447–459, 1996.
[9] F. R. Ong and K. Bouazza-Marouf, “Drilling of bone: a robustautomatic method for the detection of drill bit break-through,” ProcInst Mech Eng H, vol. 212, no. 3, pp. 209–221, 1998.
[10] P. Brett, A. Harrison, and T. Thomas, “Schemes for the identificationof tissue types and boundaries at the tool point for surgical needles,”IEEE Transactions on Information Technology in Biomedicine, vol. 4,no. 1, pp. 30–36, march 2000.
[11] P. N. Brett, D. A. Baker, R. Taylor, and M. V. Griffiths, “Controllingthe penetration of flexible bone tissue using the stapedotomy micro-drill,” Proceedings of the IMechE, Part I: J. of Systems and ControlEngineering, vol. 218, no. 5, pp. 343–351, 2004.
[12] W.-Y. Lee, C.-L. Shih, and S.-T. Lee, “Force control and breakthroughdetection of a bone-drilling system,” IEEE/ASME Transactions onMechatronics, vol. 9, no. 1, pp. 20–29, March 2004.
[13] W.-Y. Lee and C.-L. Shih, “Control and breakthrough detection of athree-axis robotic bone drilling system,” Mechatronics, vol. 16, no. 2,pp. 73–84, 2006.
[14] C. J. Coulson, A. P. Reid, and D. W. Proops, “A cochlear implantationrobot in surgical practice,” Preservation, pp. 2–4, 2008.
[15] R. Taylor, X. Du, D. Proops, A. Reid, C. Coulson, and P. N. Brett,“A sensory-guided surgical micro-drill,” Proceedings of the Institutionof Mechanical Engineers, Part C: Journal of Mechanical EngineeringScience, vol. 224, no. 7, pp. 1531–1537, 2010.
[16] V. Colla and B. Allotta, “Wavelet-based control of penetration ina mechatronic drill for orthopaedic surgery,” in IEEE Int Conf. onRobotics and Automation, vol. 1, May 1998, pp. 711–716.
[17] B. Allotta, G. Giacalone, and L. Rinaldi, “A hand-held drilling toolfor orthopedic surgery,” IEEE/ASME Transactions on Mechatronics,vol. 2, no. 4, pp. 218–229, December 1997.
[18] V. G. Kaburlasos, V. Petridis, B. Allotta, and P. Dario, “Automaticdetection of bone breakthrough in orthopedics by fuzzy lattice rea-soning (flr): The case of drilling in the osteosynthesis of long bones1,” Computer Engineering, no. 1, pp. 1–9, 1997.
[19] V. G. Kaburlasos and V. Petridis, “Fuzzy lattice neurocomputing (fln)models,” Neural Networks, vol. 13, no. 10, pp. 1145–1170, 2000.
[20] J. Savall, J. Martın, and A. Avello, “High performance linear cabletransmission,” J. Mech. Des., vol. 130, no. 6, June 2008.
[21] M. Louredo, I. Dıaz, and J. J. Gil, “Metodo de perforacion de huesoy dispositivo para llevar a cabo dicha perforacion,” Patent, 2011,P201100444.